Neuro-fuzzy logic based fusion algorithm of medical images

CT, single photon emission computed tomography (SPECT) and nuclear magnetic resonance imaging (MRI) are complementary on reflecting human information. In order to provide more useful information for clinical diagnosis, we have a need to fuse the effective information. In the pixel-level fusion between the medical images, we presented a fusion algorithm based on neuro-fuzzy logic in this paper, and utilized hybrid algorithm which mixes BP algorithm with least mean square (LMS) algorithm to train the parameters of membership function. Employ the data of medical image CT, SPECT and MRI to achieve the fusion simulation, and compare with the simulation results of BP neural network on the basis of the evaluation standards which are the standard deviation and the information entropy. By the contrast and analysis, we got the following conclusions: the fused images based on neuro-fuzzy logic not only reserve more texture features, but also enhance the information characteristics of two original images.